increased search threshold for large database

[zjing14]

I got the following Jenkins failure for large search time with a large database. Thus, I increased the allotted time threshold.

test 10/32
tC0_tA0_tB0_colMaj1_m81_n91_k67_lda90_ldb77_ldc93_ws1000000_f32
Found device gfx900 @{[64 CUs] [1500 MHz]}. Use/modify a CLHint to change.

Entering new descent.
{0.05 <<< @t=3.009e-06[s] <<< 0.1} {0 <<< @i=0 <<< 100000}
geometry : tC0_tA0_tB0_colMaj1_m81_n91_k67_lda90_ldb77_ldc93_ws1000000_f32
allotted time : 0.100000
Warmstart requested [@ rank 0] Nearest match in kernel cache:
device : gfx803' constraints : '
geometry : `tC0_tA0_tB0_colMaj1_m77_n77_k77_lda77_ldb77_ldc77_ws0_f32'
Time in get_default : 0.099062 [s]
stopping the search because allotted time has been surpassed
terminate called after throwing an instance of 'MIOpenGEMM::miog_error'
what():

[newling]

Looks like a safe change. I don't understand the details (0.1 should be enough time to find nearest match in a huge db if there is no compiling / benchmarking ... ) but I approve (0.1 was probably chosen randomly anyway).